Smartling MCP Server: Connect Your AI Chat Tools to Smartling

Smartling's MCP (Model Context Protocol) server connects Smartling to AI assistants such as ChatGPT and Claude, and to AI-powered code editors such as Claude Code, Cursor, Visual Studio Code, and OpenAI Codex. You can then translate content, look up your glossaries and translations, and manage translation jobs from your AI tool's chat, without opening the Smartling dashboard.

The easiest way to get started: Smartling's MCP server is available directly in ChatGPT and Claude.

What is MCP?

The Model Context Protocol (MCP) is an open standard developed by Anthropic. It gives AI applications a standard way to communicate with external systems. Think of MCP as a universal connector that lets AI assistants securely access resources such as databases, APIs, and other platforms.

Smartling's MCP server is a bridge between your AI tools and Smartling's translation platform. After you connect the server to your AI tool, you can use the prompt interface to start actions with the tools the server provides.

What can I use the Smartling MCP server for?

After you connect the server to your AI tool, you can:

Start actions in Smartling directly from your AI tool's chat interface
Retrieve translation assets and information from Smartling
Translate text and files instantly with machine translation or an LLM
Create and track translation jobs, and upload files to your Smartling projects
Manage glossaries and style guides, and review translation quality
Add Smartling's translation capabilities to your existing AI-powered processes

For example, you can ask your AI tool to translate content without any complex configuration. Instead of a generic translation, you can use a simple prompt to apply your Smartling assets. The output then follows your preferred terminology and style. You can also write prompts to look up existing translations or other information in your Smartling account.

In other words, instead of logging in to the Smartling dashboard, you can open your AI tool and use its prompt interface to run Smartling's MCP tools.

Which Smartling MCP tools are available?

Smartling provides MCP tools that act as building blocks. They let your connected AI tool read and use the translation resources across your Smartling account, such as glossaries, style guides, translation memory, projects, and jobs. Your AI tool calls these tools as needed based on your prompts. Once the server is connected, every available tool appears in your AI tool with a full description.

The tools fall into these groups:

Translation: Translate text and files instantly with machine translation or an LLM.
Linguistic assets: List and manage glossaries, style guides, and linguistic packages. Search translation memory and match glossary terms.
Jobs and files: Create and authorize translation jobs, add strings, upload files in batches, and download translated files.
Quality and issues: Raise issues on strings and review translation quality.
Account information: View projects, users, and invites.
Connectors: Search and translate content from connected systems such as Intercom.

For the complete list of tools, see Smartling MCP Server Tools.

How does file translation work with the Smartling MCP server?

File translation support depends on whether your MCP client has shell access. A client with shell access can run commands directly on your computer (like Claude Code or OpenAI Codex). A client without it runs only in a browser or chat window (like claude.ai on the web).

Clients with shell access (such as Claude Code and OpenAI Codex) can translate larger files and binary files, such as Word, PowerPoint, or PDF documents. Behind the scenes, the client requests a short-lived (15-minute) upload link, sends the file to Smartling through it, and then requests the translation.
- Example: Ask Claude Code to "translate this 10-page product manual Word document into French and put the translated file in the current directory." It uploads the whole file and translates it.
Clients without shell access (such as claude.ai on the web) can only send file content directly in the chat message, so only small, plain-text files are supported.
- Example: In claude.ai on the web, you can translate a short .txt snippet, but a large Word document may fail or be rejected.

To upload files into a Smartling project instead of translating them instantly (for example, to use translation memory or a human review workflow), use the job batch tools listed in Smartling MCP Server Tools. For file-based workflows that run locally with the Smartling CLI, see Smartling CLI MCP: File-Based Translation Workflows with AI Tools.

Which AI tools support the Smartling MCP server?

MCP is a relatively new technology, and not every AI tool supports it yet. You can add the Smartling connector directly from the connector directories in ChatGPT and Claude, with no configuration file. The server also works with AI-powered code editors such as OpenAI Codex, Claude Code, Visual Studio Code, Cursor, and Gemini CLI, which you configure manually.

File translation capabilities vary by client, depending on shell access. See How does file translation work with the Smartling MCP server? above.

Check the documentation for your preferred tool to confirm that it can connect to an MCP server.

How do I set up the Smartling MCP server?

Step 1: Configure your translation settings in Smartling

Linguistic assets. To get the translation output you want, keep your linguistic assets in Smartling up to date and of good quality. Review them before you start translating.

Translation provider. Unless your prompt says otherwise, the LLM uses the translation provider configured in your MT API settings.

In the Smartling dashboard top navigation, go to AI Hub > Instant MT.
Select the MT API tab.
Select the MT profile to use for translation requests made through the MCP server.
Make sure the toggle is switched on.

Glossary term insertion must be requested in your prompt, even if it is enabled on this page.

Step 2: Connect the Smartling MCP server to your AI assistant

As of August 3, 2026, Smartling MCP uses OAuth authentication. Existing token-based connections continue to work as before, but new connections require OAuth.

Choose how to connect. Most people should use Option 1: add the Smartling connector in ChatGPT or Claude. It takes a few clicks and needs no configuration file. Use Option 2 only if you work in a code editor such as VS Code, Cursor, Claude Code, or OpenAI Codex, which connect through a short configuration.

Sign-in. The Smartling MCP server uses an OAuth 2.1 login flow. Your AI tool opens a browser window where you log in to Smartling and authorize access. No token is required, and the configuration does not include an authentication header.

The Smartling MCP server has both project-level and account-level tools. Using every tool requires the Account Owner role. Users without this role can still sign in and use the tools their permissions allow.

Option 1: Add the Smartling connector in ChatGPT or Claude (no configuration file needed)

Smartling's MCP server is listed in ChatGPT and in the Claude connector directory. You add it like any other connector: find Smartling, connect it, and log in to Smartling. There is no configuration file to edit, no token to copy, and no command line to use.

ChatGPT: Open the Smartling app in ChatGPT and add it.
Claude: Open the Smartling connector in the Claude directory and add it.

If you use the Claude Desktop app, you can also add the connector from inside the app. The steps are below.

Claude Desktop supports two ways to connect to the Smartling MCP server without editing a configuration file. We recommend adding it from the Connector Directory.

Recommended: add from the Connector Directory

Open Claude Desktop and go to Settings > Connectors.
Select the Discover tab.
Search for "Smartling" and select the connector listed as by Smartling, Inc.
Click Connect to Claude.
In the browser window that opens, click Continue connecting.
Log in to Smartling and authorize access.
When you see the Connected confirmation, click Open desktop app to return to Claude Desktop.

Alternative: add a custom connector

Open Claude Desktop and go to Settings > Connectors.
Click Add custom connector.
Paste the server URL: https://mcp.smartling.com/mcp
Follow the OAuth login prompt to sign in to Smartling and authorize access.

In Team and Enterprise workspaces, only Workspace Owners can add a new custom connector to the organization. Other members can only connect to connectors an Owner has already added. For more information, see Connect to remote MCP servers.

Option 2: Configure the server manually (for developers)

Use the configuration for your AI tool below. The first time you start the server, your AI tool prompts you to log in to Smartling in your browser and authorize the connection. Check your tool's documentation for the most up-to-date MCP setup steps.

Visual Studio Code (GitHub Copilot): add this to .vscode/mcp.json

{
  "servers": {
    "smartling": {
      "type": "http",
      "url": "https://mcp.smartling.com/mcp"
    }
  }
}

Cursor: add this to ~/.cursor/mcp.json (or .cursor/mcp.json in your project)

{
  "mcpServers": {
    "smartling": {
      "url": "https://mcp.smartling.com/mcp"
    }
  }
}

Claude Code (CLI): run this command, then follow the browser login prompt

claude mcp add --transport http smartling https://mcp.smartling.com/mcp

OpenAI Codex: add this to ~/.codex/config.toml

[mcp_servers.smartling]
url = "https://mcp.smartling.com/mcp"

Gemini CLI: run this command, then follow the browser login prompt

gemini mcp add --transport http smartling https://mcp.smartling.com/mcp

Claude Desktop (fallback): manual configuration with mcp-remote

If you can't add the connector in Claude Desktop (see Option 1), you can connect through mcp-remote, a local bridge process that handles the OAuth handshake. Add the following to your claude_desktop_config.json file, restart Claude Desktop, and follow the browser login prompt.

{
  "mcpServers": {
    "smartling": {
      "command": "npx",
      "args": ["mcp-remote@latest", "https://mcp.smartling.com/mcp"]
    }
  }
}

For more information, see Connect to local MCP servers.

How do I view and manage the available MCP tools?

After the connection is established, you can view and use all available Smartling tools. In VS Code, click the Configure Tools icon in the chat window.

By default, all Smartling MCP tools are accessible. They let your AI tool read account-wide resources in Smartling, such as all of your glossaries, projects, and style guides. To turn off a tool, clear its checkbox and save the configuration.

Step 3: Create a prompt

You are ready to start prompting. After the Smartling MCP server is configured, you can start any request from your AI tool's chat interface with natural language. You can combine all available tools to get the result you want.

For example, you can ask the LLM to translate a text snippet into a specific language using Smartling machine translation.

In your prompt, you can name extra resources or linguistic assets, such as terms from your Smartling glossary. Be sure to mention that you want to use Smartling for translation.

Example prompt:

"Please translate this text into de-DE using Smartling machine translation. Use terms from the Marketing Glossary."

If your prompt doesn't say you want to use Smartling, the LLM typically returns a generic translation instead of using Smartling's MT API.

The LLM then uses the tools your request needs and returns a translation. If it needs clarification, it may ask you to confirm which Smartling resources to use. It identifies each tool it plans to run, and you click Continue to run each one.

example prompt.png

The LLM decides which tools to use when it interprets your prompt. For example, if you want terms from your Smartling glossary applied, say so explicitly. Use prompts that are as specific as possible. Smartling cannot guarantee specific translation results.

Your prompt can also use other Smartling tools to retrieve existing translation data from your account. For example, you can fetch the existing translation of a string stored in a specific translation memory or project.

example prompt -find translation.png

Important considerations

The Smartling MCP server supports two approaches to translation. The approach you choose determines whether your content is added to a Smartling project and goes through a translation workflow, with the option for human review, or is translated instantly without going through a workflow in the Smartling dashboard.

Instant translation (text and files)

When you use the machine translation tools to translate text or files, the content is translated using machine translation (MT) or a large language model (LLM) only. The content is not uploaded to a Smartling project or saved to your translation memory. Translations do not go through a translation workflow or receive human review. Smartling does not store or ingest the content you submit for instant translation.

Project-based translation (jobs, batches, and connectors)

When you use the job, job batch, string, or connector tools, your content is added to a Smartling project. You can then use translation memory and your project’s translation workflows, including workflows with human review.

These tools can also modify or delete data in your account. For example, they can authorize jobs or remove files. Review the requested actions carefully before confirming them in your AI tool.

For a complete list of available tools, see Smartling MCP Server Tools.

MCP is a fast-growing technology, and the list of Smartling MCP tools will keep evolving to support more use cases. If you are interested in using the Smartling MCP server, please feel free to contact us. Our team will be happy to assist you and evaluate how to best achieve your goals.

Get answers instantly